What you’ll learn:
- Why AI and HPC are expected to drive unprecedented market inflection.
- How government investments across all of the major countries are supporting the semiconductor industry.
As we look to break new ground in the semiconductor industry, a very bright future is unfolding. We’re entering the New Information Age, where there’s great potential to transform and redefine not only our semiconductor industry, but also the very fabric of our daily lives.
In the past, the industry’s growth and roadmap were mainly driven by one key application: processing power for computers, from mainframes to PCs, mobile devices, and smartphones. Today, we see multiple applications ramping up at the same time: artificial intelligence (AI), 5G, IoT, high-performance computing (HPC), data centers, virtual and augmented reality, and autonomous driving.
AI and HPC alone are expected to drive unprecedented market inflection, with innovation needed to meet its demand and requirements. In addition, government investments across all major countries supporting the semiconductor industry will catalyze the growth of the semiconductor market. This New Information Age—driven by digital transformation—will significantly increase the semiconductor industry's expansion and transform its landscape.
In 2025, we will see these semiconductor industry trends:
1. Foundry capacity will remain at a premium
Demand for advanced semiconductor foundry capacity will remain at an all-time high in 2025. Many leading industry analyst firms have discussed that between 2028 and 2032, there will be a shortage of leading-edge node manufacturers. In the coming year, companies need to start identifying new alternatives to semiconductor manufacturing.
2. Artificial intelligence/high-performance computing growth
2025 will continue to drive the need for AI chips. According to Gartner, AI-related semiconductors are expected to reach $97B in revenue in 2025, growing at a significantly higher rate (30%) in 2025 vs. the industry’s growth of 14%.
Developments in generative AI and large language models (LLMs) will continue to drive the deployment of high-performance processors and accelerators in data centers to support training and inference AI workloads. AI model training will drive a massive compute expansion as data centers try to keep up with demand. Even though inference is run in data centers today, we expect the next generation of edge server processors to enable inference at the edge.
Data-center-trained models will be deployed to edge servers and into IoT devices and cars. We’ll see a new generation of edge processors explicitly designed for inference calculations. Processors will continue to advance for data centers and edge servers.
3. Increased adoption of chiplets and modular type architectures
Adoption of AI across multiple applications will continue to drive the need for process customization, with chips optimized for various AI workloads. This will drive further innovation in semiconductor design.
We expect to see increased adoption of chiplets and novel architectures, development of more powerful and efficient processors, higher-bandwidth memories, and more efficient integration technologies. Innovation across the entire supply chain will be needed, from design to front- and back-end technologies, advanced interconnects, and substrates, to achieve the true potential of chiplet-based architectures. Chiplet architectures will be used not only for AI, but across other applications.
4. Growth of high-end advanced packaging and heterogeneous integration
Adopting chiplets will drive further innovation in heterogeneous integration and high-end advanced packaging technologies. While 2.5D silicon interposers will continue to be the dominant packaging platform supporting AI/HPC applications, the high cost of such interposers and supply chain availability will drive the industry to further innovate in bringing alternative and more cost-effective solutions to the market.
We will continue to see the adoption of bridge technologies, increased 3D stacking for high-bandwidth memories as well as hybrid bonding to drive finer pitch sizes, ultra-high-density fan-out adoption, development of organic interposers to enable more advanced interconnects, and advances in organic and glass substrate developments to address the increased complexity needs and larger-size packages driven by AI requirements.
5. Increased development and adoption of silicon photonics and co-packaged optoelectronics
Scaling AI will also drive the need for more advanced interconnects in the back-end networking. While copper interconnects will continue to be the primary interconnect technology for computing and die-to-die connections (short-reach connections), the high bandwidth driven by AI to connect thousands of servers across an entire data center will further drive the development and adoption of optical interconnects.
Co-packaged optoelectronics (CPO) integrate optical and electronic components into one package, enhancing the performance of the front-end networks that connect the servers with lower-power consumption at increased bandwidth density. Interest has already been raised in the industry, and we expect to see more developments and the adoption of CPOs.
Boundless Opportunities for Innovation
The semiconductor industry has always been the heartbeat of technological progress. Looking forward, we see a landscape rich with even greater growth possibilities and opportunities for innovation.
While AI/HPC is expected to have the highest growth, IoT, 5G, and autonomous driving will also contribute to the expansion of the semiconductor market and drive innovation across the entire supply chain. Advanced packaging, chiplets, and heterogeneous integration are expected to fundamentally change the semiconductor industry landscape in the coming years and unlock new frontiers of innovation.
For more details on semiconductor innovation and the latest in logic-chip fabrication, check out this podcast:
